Personalized neuromodulation in epilepsy: navigating challenges toward a convergent future

Approximately one-third of patients with epilepsy develop drug-resistant epilepsy, where conventional resective surgery faces stringent clinical limitations. Recent years have witnessed neuromodulation techniques emerging as clinical alternatives, offering new therapeutic prospects for this population, though suboptimal efficacy persists. Current neuromodulation paradigms are evolving from empirical to data-driven and model-driven approaches, progressing toward their deep integration. Data-driven methods leverage clinical multimodal data, such as stereoelectroencephalography, to identify pivotal nodes within individual epileptic networks, while model-driven approaches simulate and optimize intervention strategies through constructing a “digital twin brain model”. This review systematically evaluates clinical evidence and individualized practices of implantable closed-loop technologies, alongside advances in noninvasive deep brain modulation techniques—exemplified by temporal interference stimulation—and their multimodal synergies. Crucially, we delineate the fundamental distinctions and complementary relationships between data-driven and model-driven paradigms, proposing a unified framework centered on “digital twin brain models” to establish a closed loop from empirical observations to mechanistic simulations and treatment outcome projections. Finally, we synthesize current challenges across technological, methodological, and clinical domains, aiming to provide a theoretical foundation and actionable roadmap for personalized precision neuromodulation in epilepsy.